Pilot multi-reader study demonstrating potential for dose reduction in dual energy hepatic CT using non-linear blending of mixed kV image datasets

Objective: To determine the potential for radiation dose reduction using sigmoidally-blended mixed-kV images from dual energy (DE) hepatic CT. Methods: Multiple contrast-enhanced, DE (80 kV/140 kV) datasets were reconstructed from 34 patients undergoing clinically-indicated examinations using routine CTDI _vol. Noise was inserted in projection-space to simulate six dose levels reflecting 25-100% of the original dose. Three radiologists, blinded to dose, evaluated image preference, image quality, and diagnostic confidence (scale 1 to 5) using sigmoidally-blended, mixed-kV images, identifying the lowest acceptable dose (both image quality and confidence scores ≥4). At this lowest acceptable dose, the sigmoidal, 0.5 and 0.3 linear blended images were ranked in order of preference. Results: Radiation dose level correlated with image preference (correlation coefficients = 0.94, 0.81, 0.94). However, 82% (28/34) and 97% (33/34) of examinations corresponding to dose reductions of 45% and 30%, respectively, yielded acceptable image quality and confidence for all three radiologists. These frequencies were similar whether or not a lesion was present. Each radiologist had specific preferences between mixed-kV image display techniques (p≤0.006), with two most often preferring sigmoidally-blended images. Conclusions: There is potential for further dose reduction utilizing DE hepatic CT. Radiologist visual preference for mixed-kV images is idiosyncratic.